A COMPARATIVE ANALYSIS OF PROBABILISTIC AND MACHINE LEARNING MODELS FOR BIOLOGICAL SEQUENCE CLASSIFICATION

The human genome, encompassing about six billion nucleotides, is a comprehensive blueprint for biological processes. Understanding genetic diversity enables the identification of disease susceptibility genes, potential drug targets, and the development of diagnostic tools. Automated classification is essential in genome projects due to the vast amount of DNA sequences. This study employed probabilistic modeling (specifically Markov model and maximum likelihood estimation) along with machine learning (ML) classifiers integrated with natural language processing (NLP) to analyze intricate genomic datasets, focusing on gene family classification. We evaluated our approach on a dataset of raw Human DNA sequences, and the results demonstrated that the multinomial Naïve Bayes and multilayer perceptron classifiers achieved specificity and sensitivity exceeding 90%. The Markov model yielded results on par with ML classifiers, indicating its significance in genetics and healthcare advancements. The study’s approach, incorporating stochastic models, NLP techniques, and ML algorithms, establishes a robust framework for future genomics research and promises further insights into genetic mechanisms and therapeutic interventions.